Stable Feature Selection with Applications to MALDI Imaging Mass Spectrometry Data

TL;DR

This paper introduces a stable feature selection method using bootstrap and FDR control, analyzes the distribution of a correlation coefficient, and applies these techniques to MALDI imaging mass spectrometry data.

Contribution

It proposes a novel feature selection approach that enhances stability and provides a finite sample distribution analysis for correlation coefficients relevant to MALDI data.

Findings

Improved stability in feature selection for high-dimensional data

Finite sample distribution of Chatterjee's correlation coefficient derived

Successful application to MALDI imaging mass spectrometry data

Abstract

This paper discusses an approach, based on the subsampling boostrap and FDR control, to improve the stability of feature selection. It furthermore presents the finite sample distribution of the correlation coefficient recently proposed by Chatterjee (2020) under the setting relevant for this paper. Finally an application to matrix-assisted laser desorption/ionization (MALDI) imaging mass spectroscopy data is discussed.